Estimating construction duration of diaphragm wall using firefly-tuned least squares support vector machine

Diaphragm wall is a widely used method for excavating the foundation of buildings. Foundation construction plays a prominent role since it is a predecessor of many other activities. Therefore, estimating the diaphragm wall duration at the planning phase is a practical need of the project manager. This research proposes an artificial intelligence method, named as FLSVM, for predicting duration of diaphragm wall construction. The proposed method is developed by a fusion of the least squares support vector machine (LS-SVM) and the firefly algorithm (FA). LS-SVM is used to generalize the mapping function between the diaphragm wall duration and its influencing factors. Meanwhile, the FA method is employed for finding an appropriate set of the LS-SVM tuning parameters. Historical cases collected from construction projects in Taiwan are utilized to establish and verify the new approach. Based on the experimental results, the FLSVM model can deliver accurate forecasts since it has achieved a comparatively low prediction deviation which is <10%. This fact proves that the proposed approach is very helpful for construction managers when planning the schedule of diaphragm wall-related projects.